NetCourse 461

Introduction to Univariate Time Series Using Stata

Content:

This course provides an introduction to univariate time-series analysis that emphasizes the practical aspects that are most needed by practitioners and applied researchers. The course is written to appeal to a broad array of users, including economists, forecasters, financial analysts, managers, and anyone who encounters time-series data.
 
The course includes access to the lecture material, detailed answers to the questions posted at the end of each lecture, and access to a discussion board on which students can post questions for other students and the course leader to answer.

Course leader:
Brian Poi, senior economist at StataCorp
Gustavo Sanchez, staff statistician at StataCorp
Course length:
7 weeks (four lectures + overview of multivariate methods)
Dates:
October 8 – November 26, 2010
Prerequisites:
  • Stata 11, installed and working
  • Course content of NetCourse 101 or equivalent knowledge
  • Familiarity with basic cross-sectional summary statistics and linear regression
  • Internet web browser, such as Microsoft Internet Explorer, Netscape, or Mozilla, installed and working (course is platform independent)
Cost:
£203 + VAT

Agenda

Lecture 1: Introduction
  • Time-series data in Stata
    • Working with dates
    • Time-series operators
  • Drawing graphs
  • Simple smoothers and forecasting techniques
    • Moving averages
    • Exponential smoothers
    • Holt–Winters forecasting
Lecture 2: Descriptive analysis of time series
  • The nature of time series
    • Autocorrelation
    • White noise
    • Stationarity
  • Time-series processes
    • Moving average (MA)
    • Autoregressive (AR)
    • Mixed autoregressive moving average (ARMA)
  • The sample autocorrelation and partial autocorrelation functions
  • Introduction to spectral analysis—the periodogram
Lecture 3: Forecasting II: ARIMA and ARMAX models
  • Basic ARIMA models
    • Using ARMA processes to model series
    • Choosing the number of AR and MA terms
    • Selecting the best model from information criteria
  • Forecasting
  • Seasonal ARIMA models
  • Models with exogenous regressors—ARMAX models
  • A brief tour of intervention analysis
    • Additive outliers
    • Level shifts
There is a week-long break between lectures 3 and 4 to allow more time for those who may fall behind and for more discussion from the participants.
Lecture 4: Regression analysis of time-series data
  • Autocorrelation
    • Testing for autocorrelation
    • Obtaining Newey–West standard errors
    • More on ARMAX models
  • Seasonal effects
  • Nonstationarity and unit-root tests
  • Heteroskedasticity in time series
    • Autoregressive conditional heteroskedasticity (ARCH) models
    • Generalized ARCH (GARCH) models and extensions
    • Testing for ARCH effects
The previous four lectures constitute the core material of the course. The following lecture is optional and introduces Stata's multivariate time-series capabilities.
Bonus lecture: Overview of multivariate time-series analysis using Stata
  • Vector autoregressive (VAR) models
    • Estimating VAR models
    • Impulse–response analysis
    • Forecasting
  • Structural VARs
  • Cointegration
    • Testing for cointegration
    • Vector error-correction (VEC) models
Back to Stata homepage
Back to Timberlake Consultants
©Timberlake Consultants Limited
Last revised:02/02/2010